以下是基于 TensorFlow 的手写数字识别（MNIST 数据集）完整代码示例： ### 环境准备 ```python import numpy as np import matplotlib.pyplot as plt from tensorflow import keras from tensorflow.keras import layers from sklearn.metrics import confusion_matrix import seaborn as sns ``` ### 加载 MNIST 数据集 ```python # 导入 MNIST 数据集 (train_images, train_labels), (test_images, test_labels) = keras.datasets.mnist.load_data() # 归一化图像像素值到 [0, 1] 范围 train_images = train_images.astype('float32') / 255.0 test_images = test_images.astype('float32') / 255.0 # 添加通道维度以适配卷积神经网络输入要求 train_images = np.expand_dims(train_images, axis=-1) test_images = np.expand_dims(test_images, axis=-1) print(f"训练集形状: {train_images.shape}, 测试集形状: {test_images.shape}") ``` ### 构建模型 ```python model = keras.Sequential([ layers.Input(shape=(28, 28, 1)), # 输入层，适应灰度图像尺寸 layers.Conv2D(32, kernel_size=(3, 3), activation='relu'), # 卷积层 layers.MaxPooling2D(pool_size=(2, 2)), # 池化层 layers.Flatten(), # 展平层 layers.Dense(128, activation='relu'), # 全连接层 layers.Dense(10, activation='softmax') # 输出层，分类为 10 类 ]) model.summary() ``` ### 配置模型 ```python model.compile( optimizer='adam', # 使用 Adam 优化器 loss='sparse_categorical_crossentropy', # 多类交叉熵损失函数 metrics=['accuracy'] # 监控准确率指标 ) ``` ### 训练模型 ```python history = model.fit( train_images, train_labels, epochs=10, # 设置训练轮数 batch_size=128, # 设置批量大小 validation_split=0.1 # 划分 10% 的数据作为验证集 ) ``` ### 评估模型性能 ```python test_loss, test_acc = model.evaluate(test_images, test_labels, verbose=2) print(f"\n测试集准确率: {test_acc:.4f}") # 输出测试集上的准确率 ``` ### 绘制混淆矩阵 ```python predictions = model.predict(test_images) predicted_classes = np.argmax(predictions, axis=1) cm = confusion_matrix(test_labels, predicted_classes) plt.figure(figsize=(10, 8)) sns.heatmap(cm, annot=True, fmt="d", cmap="Blues") plt.title("Confusion Matrix") plt.xlabel("Predicted Label") plt.ylabel("True Label") plt.show() ``` --- ###